SEARCH

SEARCH BY CITATION

Keywords:

  • Human ESCs;
  • Differentiation;
  • Self-renewal;
  • Phosphatidylinositol 3-kinase

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Human ESCs (hESCs) respond to signals that determine their pluripotency, proliferation, survival, and differentiation status. In this report, we demonstrate that phosphatidylinositol 3-kinase (PI3K) antagonizes the ability of hESCs to differentiate in response to transforming growth factor β family members such as Activin A and Nodal. Inhibition of PI3K signaling efficiently promotes differentiation of hESCs into mesendoderm and then definitive endoderm (DE) by allowing them to be specified by Activin/Nodal signals present in hESC cultures. Under conditions where hESCs are grown in mouse embryo fibroblast-conditioned medium under feeder-free conditions, ∼70%–80% are converted into DE following 5 days of treatment with inhibitors of the PI3K pathway, such as LY 294002 and AKT1-II. Microarray and quantitative polymerase chain reaction-based gene expression profiling demonstrates that definitive endoderm formation under these conditions closely parallels that following specification with elevated Activin A and low fetal calf serum (FCS)/knockout serum replacement (KSR). Reduced insulin/insulin-like growth factor (IGF) signaling was found to be critical for cell fate commitment into DE. Levels of insulin/IGF present in FCS/KSR, normally used to promote self-renewal of hESCs, antagonized differentiation. In summary, we show that generation of hESC-DE requires two conditions: signaling by Activin/Nodal family members and release from inhibitory signals generated by PI3K through insulin/IGF. These findings have important implications for our understanding of hESC self-renewal and early cell fate decisions.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Human ESCs (hESCs) have significant potential as part of a general strategy to cure degenerative diseases, for the treatment of chronic injury, and as a model for early human development. A problem associated with the application of hESCs to the development of cell therapies for diseases such as diabetes has been the inability to efficiently generate definitive endoderm, the precursor cell type that gives rise to all endoderm-derived cell lineages, including those of the pancreas [1, [2], [3], [4]5]. Recent progress has been made in relation to this problem [6] by describing an efficient method for generating definitive endoderm (DE) from hESCs (hESC-DE) using recombinant Activin A in the presence of low fetal calf serum (FCS). The development of hESCs into DE by this approach closely follows the series of events associated with endoderm formation during vertebrate development.

Understanding the developmental pathways controlling DE formation at the molecular level has so far been guided by studies in Xenopus, zebrafish, and mice [7, [8]9]. Collectively, these studies suggest a conserved mechanism for mesoderm/endoderm lineage commitment involving the transforming growth factor-β (TGFβ) family member Nodal, and a common set of downstream effector molecules [7, 8]. During murine gastrulation, the process of DE formation begins with the migration of T+ (Brachyury+) endoderm precursor cells through the anterior region of the primitive streak and through the Node, located at the anteriormost position of the streak. Nodal signaling through type I (ALK4 and ALK7) and type II (ActRIIA and ActRIIB) receptors in conjunction with its coreceptor, Cripto, is central to the generation of endoderm precursor cells and for the subsequent specification of DE. Following the engagement of Nodal with its receptor, SMAD2 and SMAD3 become phosphorylated and translocate in a complex with SMAD4 to the nucleus. One outcome of these signals is to produce cells capable of forming mesoderm or DE [9]. The specification of mesoderm or endoderm is postulated to be dependent on the duration and/or magnitude of Nodal signaling [10, 11]. In the embryo, this is likely to be dictated by the position of cells in relation to the Node. High levels of Nodal signaling are required for endoderm specification, whereas lower Nodal signaling strength promotes differentiation into mesoderm [12, [13]14]. Recent evidence based on murine ESC differentiation suggests that this cell fate decision is based on the ability of a bipotential precursor, mesendoderm, to interpret different levels of Activin/Nodal signaling [15].

In this report, we identify a key role for phosphatidylinositol 3-kinase (PI3K) signaling in hESC cell fate specification and, in doing so, define an alternative method for the efficient generation of definitive endoderm from hESCs. Furthermore, we provide the first detailed gene expression profile for DE generated from hESCs. Two requirements for the generation of hESC-DE were defined: (a) Activin/Nodal signaling and (b) suppression of PI3K-dependent signaling. These findings identify the key signaling pathways required for DE formation from hESCs, which should enable us to better understand mechanisms of early lineage commitment.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Cell Culture

hESCs (BG01, BG02, H1, Cyt-25) were maintained on mouse embryo fibroblast (MEF) feeder layers [6] or on Matrigel (1:30 dilution; BD Biosciences, San Diego, http://www.bdbiosciences.com) in mouse embryo fibroblast-conditioned medium (MEF-CM). MEF-CM was produced by conditioning of mitomycin C-treated MEFs (Specialty Media, http://www.specialtymedia.com) for 24–36 hours in Dulbecco's modified Eagle's medium/Ham's F-12 medium (DMEM/F12), high (20%) Knockout Serum Replacement (KSR) (Gibco, Grand Island, NY, http://www.invitrogen.com), 8 ng/ml human recombinant fibroblast growth factor 2 (Fgf2) (R&D Systems Inc., Minneapolis, http://www.rndsystems.com), 1 mM nonessential amino acids, l-glutamine, penicillin/streptomycin, 0.1 mM 2-mercaptoethanol. Cultures were routinely passaged with collagenase (type IV, 200 U/ml; Gibco) and split 1:4 to 1:8 every 5–6 days. Normal karyotype of hESCs was routinely confirmed by analysis of chromosome spreads from KaryoMax colcemid (Gibco)-treated cells (Medical College of Georgia Cytogenetics Unit). Differentiation was performed by plating collagenase/trypsin passaged cells at a density of 1–5 × 104 cells per cm2 on Matrigel in MEF-CM, 8 ng/ml Fgf2, 20% KSR. LY 294002 was maintained in cultures over the duration of differentiation experiments at the indicated concentrations. Recombinant Activin A, Chordin, Follistatin, Nodal, Noggin, and Lefty-A were from R&D Systems. Other reagents and sources were as follows: AKT1-II (Calbiochem, San Diego, http://www.emdbiosciences.com), SB-43152 (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), LY 294002 (Biomol, Plymouth Meeting, PA, http://www.biomol.com), purified insulin (bovine pancreas), and insulin-like growth factor 1 (IGF1) (Sigma-Aldrich).

Antibodies, Western Blot Analysis, Cell Immunostaining, and Flow Cytometry

hESC lysates were prepared using RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, 0.5% SDS) and then resolved on 8%–10% polyacrylamide gels. Following transfer to nitrocellulose membranes, filter (0.45 μm; Bio-Rad) blots were probed with primary then secondary antibodies and then developed using Amersham ECL reagents (Amersham Biosciences, Piscataway, NJ, http://www.amersham.com). The following antibodies were used in this study: rat anti-human SOX17 antibodies [6], glycogen-synthase kinase (GSK)3β (610202; BD Biosciences), T/Brachyury (AF2085; R&D Systems), fluorescein isothiocyanate-conjugated CD9 (CBL 162F; Cymbus, Southampton, U.K., http://www.researchd.com), OCT4 (N-19; Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com), and cyclin-dependent kinase 2 (CDK2) (M2; Santa Cruz Biotechnology). The following antibodies were from Cell Signaling Technology (Beverly, MA, http://www.cellsignal.com): phospho-AKT1S473 (9271); phospho-GSK3βS9 (9336); phospho-S6S235/236 (2211); phospho-S6K/p70T389 (9205); SMAD2 and SMAD3 (3102); phospho-SMAD2 and phospho-SMAD3S465/467 (3101); and phospho-SMAD1, phospho-SMAD5S463/465, and phospho-SMAD8S426/428(9511). Flow cytometry was performed as described previously [6] using an anti-CXCR4 antibody (MAB-170; R&D Systems).

Transcript Analysis by Quantitative Polymerase Chain Reaction

RNA was prepared using Qiagen RNeasy Mini Kits. Chromosomal DNA was removed using RNase-free DNase (Qiagen, Hilden, Germany, http://www1.qiagen.com). cDNA was prepared using the Superscript First Strand synthesis system (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) using 2 μg of total RNA. Target mRNAs were assayed using TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) (supplemental online Table 1), supplemented with Universal PCR Master Mix (Applied Biosystems). Polymerase chain reactions (PCRs) were performed on an Applied Biosystems ABI 7700 sequence detector.

Gene Expression Profiling by Affymetrix Microarray Analysis

Two parallel experiments were performed to generate DE by alternate approaches. First, manually passaged Cyt-25 hESCs maintained on MEF feeder layers were differentiated into DE by addition of recombinant Activin A in the presence of low FCS for 4 days, as described previously [6]. The second method of generating DE involved addition of LY 294002 to BG01 collagenase/trypsin passaged hESCs cultured on Matrigel in the presence of MEF-CM for 4 days (above). Microarray data were collected at Expression Analysis, Inc. (Durham, NC, http://www.expressionanalysis.com) using the GenChip Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA, http://www.affymetrix.com). The quality and quantity of each RNA sample was assessed using a 2100 BioAnalyzer (Agilent Technologies, Palo Alto, CA, http://www.agilent.com). The target was prepared and hybridized according to the Affymetrix Technical Manual. Total RNA (10 μg) was converted into cDNA using Reverse Transcriptase (Invitrogen) and a modified oligo(dT)24 primer that contains T7 promoter sequences (GenSet, Farmingdale, NY, http://www.enzdoio.com). After first-strand synthesis, residual RNA was degraded by the addition of RNase H, and a double-stranded cDNA molecule was generated using DNA polymerase I and DNA ligase. The cDNA products were incubated with T7 RNA polymerase and biotinylated ribonucleotides using an In Vitro Transcription kit (Enzo Diagnostics). The hybridization cocktail was denatured at 99°C for 5 minutes, incubated at 45°C for 5 minutes and then injected into a GeneChip cartridge. The GeneChip array was incubated at 42°C for at least 16 hours in a rotating oven at 60 rpm. GeneChips were washed with a series of nonstringent (25°C) and stringent (50°C) solutions containing variable amounts of MES, Tween 20, and sodium chloride/sodium phosphate/EDTA. The microarrays were then stained with streptavidin phycoerythrin, and the fluorescent signal was amplified using a biotinylated antibody solution. Fluorescent images were detected in an GeneChip Scanner 3000, and expression data were extracted using the GeneChip Operating System v 1.1 (Affymetrix). All GeneChips were scaled to a median intensity setting of 500. CEL files were normalized using probe quantile normalization, and signal values were computed using RMA algorithm with REDI analysis to reduce the impact of invariant probes.

Kidney Capsule Assays, Histology, and Immunostaining

Untreated hESCs or hESCs treated for 4 days with LY 294002 were collagenase treated to generate cell aggregates (approximately 50 cells per aggregate), washed in warm medium, and then gently resuspended in 2 ml of DMEM/F12, 10% FCS and left overnight at 37°C in 10% CO2 to facilitate further aggregation. Approximately 2.5 × 106 cells were injected into the kidney capsule of 5-week-old male severe-combined immunodeficient (SCID)-beige mice. Six weeks after transplantation, mice were sacrificed, and then kidneys were removed and fixed in 10% formalin. Following fixation, kidneys were embedded in paraffin wax, sectioned, and mounted onto glass slides in preparation for immunostaining. After mounting, slides underwent deparaffinization, rehydration, and heat-induced epitope retrieval with Trilogy (CMX-833; Cell Marque, Hot Springs, AZ, http://www.cellmarque.com). Slides were stained using the alkaline phosphatase Vectastain ABC System (AK-5002; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com) and Vector Red substrate (SK-5100; Vector Laboratories). The following antibodies were used for immunocytochemistry: albumin (A0433; Sigma-Aldrich), α-fetoprotein (AFP) (CMC700; Cell Marque), gastrin (CMC106; Cell Marque), hepatocyte-specific factor (CMC773; Cell Marque), liver fatty acid binding protein (LFABP) (RDI-FABP-L2E3; Fitzgerald Industries, Concord, MA, http://www.fitzgerald-fii.com), thyroid transcription factor 1 (TTF1) (CMC572; Cell Marque), and villin (CMC833; Cell Marque). For fluorescence immunocytochemistry, the following secondary antibodies were used: Alexa Fluor 488 goat anti-mouse IgG (A11001; Molecular Probes Inc., Eugene, OR, http://probes.invitrogen.com), Alexa Fluor 594 goat anti-rabbit IgG (A11012; Molecular Probes). Glycogen granules in grafted cells were detected by staining with a Periodic Acid-Schiff kit (87007; Richard-Allan Scientific, Kalamazoo, MI, http://www.rallansci.com).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

hESCs Differentiate When PI3K Signaling Is Inhibited

hESCs can be maintained as a stable self-renewing population under feeder-free conditions when cultured in the presence of MEF-CM, Fgf2, and high levels (20%) of KSR or FCS [16, 17]. While screening for small molecule inhibitors that could have an impact on signaling pathways involved with hESC self-renewal under these conditions, we found that several inhibitors of PI3K signaling, including LY 294002 and the AKT1 inhibitor AKT1-II, caused a marked change in cell morphology that was associated with downregulation of E-cadherin (Fig. 1A; supplemental online Fig. 1; data not shown). This coincided with a downregulation of ESC markers such as CD9 and OCT4 and upregulation of T and SOX17 (Fig. 1B–1D). Although higher concentrations (>50 μM) of LY 294002 caused some cell death in a concentration-dependent manner, consistent with it having a role in hESC survival [18], levels used in these studies had only a modest effect on colony morphology and cell viability (Fig. 1; data not shown). These observations point toward a role for PI3K in hESC self-renewal, as has been shown previously for murine ESCs [19, [20], [21]22].

thumbnail image

Figure Figure 1.. hESCs differentiate when phosphatidylinositol 3-kinase (PI3K) signaling is inhibited. (A): Collagenase/trypsin-passaged BG01 hESCs were plated on Matrigel in MEF-CM and Fgf2 and grown in the presence or absence of LY 294002 for 4 days. Bright-field images were taken at the indicated magnifications. (B): CD9 expression assayed by flow cytometry. Untreated hESCs and LY 294002-treated cells (4 days) were stained with a fluorescein isothiocyanate-conjugated anti-CD9 antibody (control, isotype-matched control antibody). (C): Immunocytochemistry of paraformaldehyde-fixed cells. Left panel, untreated BG01 hESCs grown in MEF-CM, Fgf2, and 20% KSR; middle panel, differentiated hESCs grown for 4 days in the absence of MEF-CM and Fgf2; right panel, hESCs grown with MEF-CM, Fgf2, and 20% KSR and treated with LY 294002 for 4 days. Cells were stained as indicated with DAPI and probed simultaneously with anti-CD9 and antiSOX17 antibodies. Magnification, ×20 objective. (D): Immunoblot analysis (20 μg of total protein per track) of T, CDK2, and OCT4 protein levels in untreated hESCs (−) or during differentiation of LY 294002-treated hESCs (BG01) over 6 days. (E): Q-PCR analysis of marker transcripts associated with formation of the three germ layers (ectoderm, mesoderm, and endoderm), extraembryonic endoderm, hESCs, and mesendoderm stages of differentiation following treatment with LY 294002 (50 μM) for up to 5 days. Experiments were performed in quadruplicate and expressed as the fold increase over untreated hESCs ± SEM. (F): Temporal changes in gene expression associated with the generation of hESC-DE (A) (supplemental online Fig. 2) paralleled that seen in early embryonic development. Abbreviations: CDK, cyclin-dependent kinase; CM, conditioned medium; d, day(s); DAPI, 4,6-diamidino-2-phenylindole; DE, definitive endoderm; Fgf, fibroblast growth factor; GSC, goosecoid; hESC, human ESC; PS, primitive streak stage of development when mesendoderm was formed.

Download figure to PowerPoint

We next asked whether differentiation promoted by PI3K inhibition was associated with the formation of any specific cell lineage. Following treatment with LY 294002, we found increases in transcript levels specifically associated with mesendoderm (T and MIXL1), followed by definitive endoderm, such as FOXA2, goosecoid (GSC), GATA4, GATA6, and SOX17 (Fig. 1E, 1F; supplemental online Figs. 2, 3). No consistent increases in transcript levels associated with extraembryonic endoderm (SOX7, THBD, and AFP), mesoderm (TBX6, MEOX1, FLK1, and MYOD1), or ectoderm (PAX6, SOX1, and ZIC1) were observed throughout these experiments (Fig. 1E; supplemental online Figs. 2, 3). Since many of the markers for DE are also elevated in extraembryonic endoderm, discrimination between these lineages is often problematic. To help resolve this issue, the absence of significant amounts of extraembryonic endoderm in LY 294002-treated cultures was confirmed by showing that <1% of cells were positive for AFP (visceral endoderm) and THBD (parietal endoderm; supplemental online Fig. 4). Overall, this indicates that only low levels of extraembryonic endoderm are being produced in these experiments. This will be addressed in further detail below.

Inhibition of PI3K signaling by LY 294002 was confirmed by probing cell lysates with phosphospecific antibodies that recognize the activity status of downstream PI3K signaling effectors, such as AKT1, GSK3, and ribosomal S6 kinase (S6K) and its substrate, ribosomal S6 protein (Fig. 2). These results showed that LY 294002 effectively inhibited PI3K signaling within 6–12 hours, as shown by a collapse in phosphorylation of AKT1S473, pS6S235/236, pS6K/p70T389, and GSK3βS9 (Fig. 2B, 2C). We conclude that inhibition of the canonical PI3K signaling pathway is sufficient to promote hESC differentiation without the addition or subtraction of any factors involved in self-renewal.

thumbnail image

Figure Figure 2.. LY 294002 inhibits PI3K signaling in BG01 hESCs. (A): The canonical PI3K signaling pathway. Points at which the PI3K pathway can be targeted by chemical inhibitors such as LY 294002 (PI3K) and AKT1-II (AKTI) are indicated. (B): Changes in AKT1 (AKT1S473) and ribosomal S6 (S6S235/236) protein activity in response to LY 294002 treatment (6 hours) were determined using phosphospecific antibodies in immunoblot assays (20 μg of whole cell protein). An antibody recognizing pan-GSK3β was used to evaluate relative protein loading. (C): LY 294002 treatment (6 and 12 hours) caused a rapid collapse in the activity of downstream effectors for PI3K. Activity of the canonical PI3K pathway was evaluated by analysis of downstream effectors, GSK3β and S6 kinase using phospho (activity)-specific antibodies by immunoblot analysis (20 μg of total protein). CDK2 was used as a protein loading control. Abbreviations: hESC, human ESC; hr, hour(s); IGF, insulin-like growth factor; PI3K, phosphatidylinositol 3-kinase.

Download figure to PowerPoint

Suppression of PI3K Promotes Differentiation of hESCs into Mesendoderm and Then Definitive Endoderm

When used in combination with other markers, such as SOX17, CXCR4 is a reliable marker for DE generated from murine and human ESCs [6, 23]. To finally establish that we generated a cell population predominantly enriched for DE following treatment of hESCs with LY 294002, we evaluated the CXCR4 and SOX17 status of our cultures by immunostaining and flow cytometry. Typically, LY 294002 treatment generated a population of cells that were 70%–80% SOX17+ and >80% positive for CXCR4 by day 5 (Fig. 3A–3C). More than 95% of SOX17+ cells in these cultures also stained positive for FOXA2 (data not shown). A parallel increase in CXCR4 and SOX17 mRNA immediately following the collapse of T transcripts levels (Fig. 3D) strongly argues that our cultures consist predominantly of definitive endoderm and not extraembryonic endoderm, since DE is directly derived from mesendoderm. CXCR4, SOX17, and GSC transcripts were also elevated in AKT1-II-treated hESCs (supplemental online Fig. 5). In contrast, THBD and AFP mRNAs did not increase significantly. These data reinforce earlier results showing that inhibition of PI3K signaling promotes differentiation of hESCs into DE under our culture conditions.

thumbnail image

Figure Figure 3.. Suppression of phosphatidylinositol 3-kinase (PI3K) promotes differentiation of BG01 hESCs into a CXCR4+, SOX17+ cell population. (A): hESCs and LY 294002-treated cells (5 days) were subject to immunostaining by probing for OCT4 and SOX17 and by staining with DAPI. (B): The percentage of OCT4+ and SOX17+ cells at each time point was determined by immunocytochemistry. At least 200 cells were scored in each sample. Data are shown as the standard error of the mean for three independent experiments. (C): Untreated hESCs (black) or hESCs treated for 5 days with LY 294002 (red, as in [A]) were probed with an anti-CXCR4 antibody and analyzed by flow cytometry. Per sample, 1 × 105 cells were analyzed. (D): Quantitative polymerase chain reaction analysis of T, CXCR4, and SOX17 mRNAs following treatment with LY 294002 over a 96-hour period. All reactions were performed in triplicate. Data are shown as the standard error of the mean. Abbreviations: DAPI, 4,6-diamidino-2-phenylindole; hESC, human ESC.

Download figure to PowerPoint

Authentic DE is formed through a bipotential mesendoderm (T+) intermediate [6, 15, 23]. To confirm that our SOX17+ cells transitioned through a T+ precursor state, we show overlapping expression of SOX17 and T protein at an intermediate stage of differentiation (Fig. 4). Approximately 25% of cells assayed on day 3 following addition of LY 294002 stained double-positive for SOX17 and T protein. Although the fraction of double-positive cells may seem low, if all DE must transition through a mesendoderm intermediate, the results are consistent with our observations that there is only partial temporal overlap in the accumulation of these markers (Figs. 1E, 3, 4). For example, the time when the number of SOX17+ cells begins to increase (day 3) coincides with a marked decline in the fraction of T+ cells (Figs. 3, 4). The data argue, based on what is known about the development of DE from mesendoderm [6, 23], that the CXCR4+ SOX17+ population in our cultures is being generated through a T+ intermediate cell type, consistent with definitive and not primitive endoderm being formed.

thumbnail image

Figure Figure 4.. LY 294002-treated SOX17+ cells transition through a T+ mesendoderm-like precursor. (A): Human ESCs (BG01) treated with LY 294002 for 3 days were analyzed by immunocytochemistry. Cells were probed with T and SOX17 antibodies and stained with DAPI as indicated. T+ and SOX17+ double-labeled cells stained yellow. (B): The percentage of T+ (gray bars) and double-stained cells (T+ and SOX17+; broken line) were plotted over 4 days following treatment with LY 294002. At least 200 cells were scored in each sample. Data are shown as the average of two independent experiments. (C): The percentage of SOX17+ (white bars) and double-stained cells (T+ and SOX17+; broken line) were plotted over 4 days following treatment with LY 294002. Samples were analyzed as in (B). Abbreviation: DAPI, 4,6-diamidino-2-phenylindole.

Download figure to PowerPoint

Overall, these results indicate that inhibition of PI3K promotes hESC differentiation by promoting the formation of mesendoderm and then DE but not significant levels of other lineages. Based on gene expression profiling and immunostaining, the cell type generated is indistinguishable from DE formed by treatment of hESCs with Activin A in the presence of low serum [6]. The generation of DE-like cells using PI3K inhibitors was reproducible between multiple hESC cell lines (H1, BG01, and BG02; supplemental online Fig. 3) and was independent of passage number, since similar results were obtained from cultures ranging from passage 20 to passage 80 (data not shown).

Expression Profiling of hESC-DE and Comparisons with Embryonic DE

Since the initial methods for generation of hESC-DE have only been recently reported [6], little is known about the overall gene expression profile of this cell type. To establish a more detailed molecular description of definitive endoderm and to confirm that PI3K inhibition generated a similar DE population, we expression-profiled hESC-derived DE generated by two independent methods: first, hESCs grown in MEF-CM treated with LY 294002 for 4 days (as described in this report); and second, Activin A-treated cells in low FCS [6]. Samples from replicate experiments were then used to probe Affymetrix GeneChips. The strategy to compare DE generated by the two approaches was based on the premise that it would eliminate changes in gene expression associated with specific culture conditions and would allow for the identification of a common set of genes characteristic of hESC-DE. Comparisons were made between gene expression levels in two different populations of DE and the hESCs from which they were derived. Totals of 236 and 232 transcripts increased more than 10-fold for LY 294002 and Activin A-treated cells, respectively. Seventy-five of these genes were common to both sets of data (Table 1). All of the established marker transcripts for DE were among this set, including SOX17, GSC, CXCR4, GATA4, GATA6, FOXA2, MIXL1, CER1, and HHEX, confirming that the two populations are closely related. Additional markers identified in this study further define the DE state.

Table Table 1.. Transcripts upregulated in hESC definitive endoderm >10-fold in microarray analysis
Thumbnail image of

Although it is difficult to directly compare the core set of hESC-DE genes that we identify in this report with those enriched in mouse embryonic mouse endoderm [24] or that derived from mESCs [23], several points can be made. In contrast to recent reports describing the generation of DE from mESCs, we did not observe significant levels of T protein expression in hESC-DE, consistent with DE representing a poststreak stage of endoderm development (Fig. 1F). In addition, several genes upregulated in hESC-DE have previously been identified in mouse embryonic endoderm (e7.5) [24]. These include chemokine orphan receptor 1, Dickkopf-1, Eyes Absent 2, and LGR5 (orphan G-protein-coupled receptor). In summary, this analysis defines the core set of genes associated with the hESC-DE state and identifies potential regulators of endoderm development. Moreover, it confirms the generation of a similar DE-like cell type from hESCs using two independent approaches.

Activin/Nodal Can Specify DE from hESCs Only When PI3K Signaling Is Low

A previous report describing the production of definitive endoderm from hESCs indicates that recombinant Activin A can only specify DE in low FCS [6]. With this information, we reasoned that PI3K may block differentiation of self-renewing hESCs into DE by antagonizing Activin/Nodal like activities in our cultures. Although no recombinant Activin A was included in our hESC cultures, such activities have been described in MEF-CM [25]. Release of antagonistic signals that block specification by Activin would then establish conditions compatible with differentiation into mesendoderm and then DE or mesoderm. To test these ideas, we set out to define the factors and signaling pathways that specify hESC differentiation in response to PI3K inhibition.

We first approached the problem by asking whether any TGFβ family members in our cultures were required for differentiation in conjunction with reduced PI3K signaling using inhibitors of the TGFβ pathway. Although inhibitors of bone morphogenetic protein signaling such as Chordin and Noggin had no effect on the ability of LY 294002 to promote differentiation into DE, inhibitors of Nodal/Activin (Follistatin and SB-43152) blocked the effect (Fig. 5A). This suggested that Activin signaling is necessary for differentiation and that this has an absolute requirement for low PI3K signaling. Since SOX17 upregulation was refractory to addition of recombinant Lefty-A, it is most likely that Activin in MEF-CM drives DE formation in our system, since Lefty blocks the Nodal coreceptor, Cripto, but not Activin-dependent receptor signaling [26]. These results indicate that Activin/Nodal signals from MEF-CM specify DE, but only when PI3K activity is low. This raises the possibility that PI3K suppresses the ability of Activin/Nodal-like activities to specify cell fate outcomes in hESCs and hence plays a role in self-renewal by imposing a differentiation blockade. These findings are consistent with the known role of Activin/Nodal-like activities in specifying cell lineages such as DE in the embryo and in hESCs, but for the first time point toward a role for PI3K in antagonizing the ability of these factors to specify cell fate outcomes.

thumbnail image

Figure Figure 5.. Activin/Nodal specifies definitive endoderm (DE) when phosphatidylinositol-3-kinase signaling is reduced. (A): Activin-like activities work in collaboration with LY 294002 to drive formation of DE. LY 294002-dependent SOX17 expression is suppressed by inhibitors of Activin signaling but not bone morphogenetic protein signaling or Cripto-dependent signaling. SOX17 expression was evaluated by quantitative polymerase chain reaction (Q-PCR) 4 days after addition of LY 294002 (50 μM) to BG01 hESCs grown in MEF-CM, 8 ng/ml Fgf2, high (20%) KSR in the presence or absence of Chordin (500 ng/ml), Follistatin (500 ng/ml), Lefty-A (500 ng/ml), Noggin (500 ng/ml), or SB-43152 (10 μM). Assays were performed in triplicate and are shown ± SEM. (B): MEF-CM can be substituted by recombinant Activin A or Nodal to promote LY 294002-dependent DE formation. SOX17 mRNA levels (fold increase over untreated) were evaluated by Q-PCR under various conditions over 4 days; shown are LY 294002 (50 μM), Activin A (100 ng/ml), and Nodal (1 μg/ml). UCM was composed of Dulbecco's modified Eagle's medium/Ham's F-12 medium supplemented with 4 ng/ml Fgf2, high (20%) KSR. (C): Changes in AKT1 activity under different DE differentiation conditions. hESCs were grown in unconditioned medium in low (2%) KSR plus 100 ng/ml Activin A or in MEF-CM containing high (20%) KSR and LY 294002 (50 μM). AKT1 activity was evaluated by probing cell lysates (20 μg of protein) using an activity-specific antibody directed against serine 473 (S473). Abbreviations: d, day(s); hESC, human ESC; KSR, knockout serum replacement; MEF-CM, mouse embryo fibroblast-conditioned medium; UCM, unconditioned medium.

Download figure to PowerPoint

To investigate the instructive signals required for DE formation in greater detail, we reconstituted the hESC-DE differentiation system in the absence of MEF-CM by supplementing unconditioned medium with recombinant Activin A or Nodal. The main question was to ask whether Activin/Nodal, under conditions of low PI3K, was sufficient to drive differentiation of hESCs. The approach was to remove MEF-CM from cultures and replace this with unconditioned medium plus 20% KSR, supplemented with additional factors to determine the minimal signaling requirements to promote differentiation. In the absence of exogenously added Nodal or Activin A, unconditioned medium plus 20% KSR failed to support differentiation into DE following addition of LY 294002 (Fig. 5B; data not shown). Supplementation with recombinant Nodal or Activin A, however, stimulated SOX17 expression by over 400-fold in an LY 294002-dependent manner, comparable to the effects seen in the presence of MEF-CM (Fig. 5B). These findings confirm our earlier results (Fig. 5A) indicating that Activin/Nodal specifies DE in our system but only under conditions where PI3K signaling is suppressed. This observation allowed us to focus on the relationship between PI3K and Activin/Nodal signaling in terms of their respective roles in self-renewal and differentiation.

Established methods to differentiate human and murine ESCs into DE require Activin A and low serum conditions [6, 23], in contrast to self-renewal, which requires elevated FCS/KSR. Although it was previously unclear why low FCS/KSR is absolutely required for efficient DE specification following the addition of recombinant Activin A, we reasoned that it could be related to a requirement for decreased PI3K signaling. To confirm this, we evaluated AKT1 activity as a readout for PI3K signaling and found that under conditions that support differentiation of hESCs into DE (2% KSR + Activin A), PI3K activity was significantly reduced (Fig. 5C). As expected, LY 294002 completely abolished AKT1 phosphorylation. This is consistent with our hypothesis that low PI3K signaling is a requirement for DE specification from hESCs.

Insulin/IGF Promotes hESC Self-Renewal Through the PI3K Pathway by Blocking Differentiation into DE

Consistent with previous findings [6], we observed that Activin A can only promote differentiation of hESCs into DE when KSR is present at low levels (<2%; Fig. 6). Together with our findings that low PI3K signaling is necessary for Activin A to specify DE, this suggested that factors in FCS/KSR drive PI3K signaling, thereby blocking differentiation.

thumbnail image

Figure Figure 6.. Insulin and IGF1 suppress definitive endoderm (DE) differentiation. (A): Insulin antagonizes the ability of Activin A to drive human ESC (hESC)-DE (BG01) formation in low-knockout serum replacement (KSR) unconditioned medium (UCM). Quantitative polymerase chain reaction (Q-PCR) analysis of marker transcripts was performed in triplicate; results are shown ± SEM. Conditions were as follows: UCM with 20% KSR (20K); 20% KSR UCM with 100 ng/ml Activin A (20K+A); 2% KSR UCM, 100 ng/ml Activin A (2K+A); 2% KSR UCM, 100 ng/ml Activin A, 1 μg/ml insulin (2K+AI). (B): Effects on insulin (0–18 μg/ml) and IGF1 (0 and 25 ng/ml) on DE differentiation in the presence of Activin A (100 ng/ml) monitored by Q-PCR analysis of SOX17 mRNA. Data are shown as the fold increase over untreated cultures. Assays were performed in triplicate; results are shown ± SEM. Manually passaged Cyt-25 hESCs were cultured on MEFs in the presence of 2% FCS. Abbreviations: A, 100 ng/ml Activin A; AFP, α-fetoprotein; AI, Activin plus insulin; FCS, fetal calf serum; GSC, goosecoid; Igf, insulin-like growth factor; K, knockout serum replacement; NF, no factor added.

Download figure to PowerPoint

Since factors in FCS/KSR antagonize the ability of Nodal/Activin to specify DE, we sought to identify the agonist responsible for PI3K signaling that blocked differentiation. Since insulin and IGF are present at biologically active levels in KSR and FCS, respectively, and since both are well-established agonists of PI3K-dependent signaling, this suggested that LY 294002 promotes DE formation by inhibiting effectors of insulin/IGF. This was confirmed by showing that addition of insulin or IGF to hESCs, grown in unconditioned medium with low KSR or FCS, severely reduced the ability of Activin A to upregulate DE markers such as SOX17, MIXL1, and GSC (Fig. 6A, 6B; data not shown). This is consistent with the hypothesis that insulin/IGF blocks hESC differentiation by signaling through PI3K. The general applicability of insulin/IGF as a DE antagonist has been confirmed in multiple hESC lines (BG01, BG02, H1, and Cyt-25), cultured by manual or collagenase passaging under feeder-dependent and feeder-free (Matrigel) conditions.

hESC-DE Differentiates into Specialized Endoderm Lineages In Vivo

To determine whether hESC-DE cells generated by treatment with LY 294002 were functionally competent to differentiate into more specialized cell types of endodermal origin, we transplanted them into the kidney capsule of SCID-beige mice [6]. A general prediction was that for hESC-DE to be considered as an authentic endoderm progenitor it must be competent to generate multiple endoderm cell types and be more predisposed to generate these cell types than undifferentiated hESCs. After 6 weeks, engrafted cells (Fig. 7A) were fixed with formalin for immunostaining or frozen for subsequent RNA extraction and quantitative polymerase chain reaction (Q-PCR) analysis. Analysis of several endoderm antigens associated with liver (AFP, albumin, hepatocyte-specific antigen, and LFABP), thyroid/lung (TTF1), and intestinal epithelia (villin and gastrin) were detected extensively throughout all DE grafts examined (Fig. 7B, 7C). Clusters of LFABP+ AFP+ cells also stained positive for glycogen granules using Schiff's reagent (data not shown), consistent with the production of liver lineages. Only one of nine grafts from untreated hESCs, however, showed any reactivity with these antibodies or to Schiff's reagent, and even then, it was limited compared with hESC-DE grafts.

thumbnail image

Figure Figure 7.. BG01 human ESCs (hESCs)-definitive endoderm (DE) are predisposed to generate specialized cell types of endoderm origin. (A): Engrafted tissue derived from hESC-DE (+LY 294002) 6 weeks after transplantation into the kidney capsule of SCID-beige mice. Control represents a mock-injected kidney receiving medium only. (B): Grafts produced 6 weeks after transplantation of hESC-DE into the kidney capsule were formalin-fixed, paraffin-embedded, and then sectioned and subjected to immunohistochemistry by probing with TTF1 (lung/thyroid marker; magnification, ×20 objective), AFP (liver; magnification, ×20), villin (gastrointestinal epithelia, 20 x magnification), gastrin (stomach, magnification, ×40), HSA (magnification, ×40), and LFABP (magnification, ×20 objective). (C): Immunofluorescence staining of engrafted tissue derived from hESC-DE. LFABP (green), albumin (red), DAPI (blue), and superimposed LFABP/albumin (yellow) are shown (magnification, ×40 objective). (D): Quantitative polymerase chain reaction (Q-PCR) analysis of whole engrafted tissue taken 6 weeks after transplantation of hESC-DE or hESC (BG01) in to the kidney capsule. FABP1 and albumin mRNA was analyzed by Q-PCR in quadruplicate, (n = 5; ±SEM). Abbreviations: AFP, α-fetoprotein; DAPI, 4,6-diamidino-2-phenylindole; FABP, fatty acid binding protein; HSA, hepatocyte-specific antigen; LFABP, liver fatty acid binding protein; TTF, thyroid transcription factor.

Download figure to PowerPoint

Comparison of mRNA levels in whole grafts indicated that hESC-DE expressed approximately 10-fold more albumin and FABP1 mRNA than equivalent grafts produced by transplantation of hESCs (Fig. 7D). This was unlikely to be due to the presence of extraembryonic endoderm since TM and SPARC remained low, as was the case in the implanted hESC-DE. Furthermore, markers for mesoderm (hemoglobin, collagen 1A1 and 2A1, MYOD1, and NKX2.5) and ectoderm (GFAP, MAP2, NKX2.2, OLIG2, PAX9, and synaptophysin) were not enriched in DE-transplanted material after 6 weeks (data not shown). These results indicate that hESC-DE was more predisposed to form endoderm lineages than hESCs, consistent with the transplanted cells being functional DE.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

hESCs can be efficiently differentiated into DE by the addition of exogenous Activin A but only in the presence of low FCS/KSR [6]. Initially, it was difficult to reconcile why addition of PI3K inhibitors should recapitulate this differentiation program, but subsequently we were able to conclude several things. First, hESCs grown on MEF feeders or MEF-CM are already exposed to levels of Activin A sufficient to promote specification of DE. Second, hESCs are unresponsive to Activin A in MEF-CM because of an antagonistic effect mediated by PI3K signaling. KSR (insulin) and FCS (IGF) activate PI3K in hESCs and were shown to be potent inhibitors of Activin-dependent differentiation. Previous work has shown that Nodal/Activin signaling thresholds dictate whether mesendoderm moves toward mesoderm or endoderm cell fates. In our experiments, Activin signaling is sufficiently high to specify DE when PI3K activity is low. We predict that mesendoderm would differentiate into mesoderm, however, if Activin levels in our system were reduced. It is possible that the inhibitory effects of PI3K may extend to other differentiation pathways, but this was not addressed by our work here.

Several observations indicate that the cell type generated by treatment of hESCs with LY 294002 is authentic, definitive endoderm. Marker analysis at the protein and RNA levels is consistent with the formation of DE and excludes the possibility that significant levels of extraembryonic endoderm or other lineages are being produced. Microarray analysis indicates that the cell types generated by Activin A and LY 294002 treatment, under conditions of low PI3K activity, are very similar. Furthermore, as is the case for Activin-treated cells, DE generated by treatment with PI3K inhibitors first transition through a T+ stage before expressing markers such as SOX17 and CXCR4. This is reminiscent of the development of DE from a primitive streak intermediate in the vertebrate embryo. As described for the formation of DE from hESCs previously, we saw a decline in E-cadherin expression following LY 294002 treatment (supplemental online Fig. 1), consistent with differentiation being associated with an epithelial to mesenchymal transition similar to that in the primitive streak [6]. This is associated with a reciprocal change in N-cadherin expression (supplemental online Fig. 1), as has been described previously [6]. It is unclear why there is a difference between the E-cadherin status of DE derived from murine and human ESCs [6, 23]. Finally, as was recently described for Activin/low FCS-derived DE [6], endoderm produced by treatment with PI3K inhibitors can differentiate into more specified endoderm lineages in a kidney capsule assay. The production of similar endoderm populations by different approaches was initially surprising, but our work has resolved this paradox by establishing an interaction between the same two signaling pathways, established under different culture conditions.

There are several critical parameters that influence the ability of LY 294002 to drive DE formation under the culture conditions described in this report. First, although LY 294002 at concentrations between 20 and 50 μM works effectively, below this level, the efficiency of DE formation diminishes in a concentration-dependent manner (data not shown), presumably because PI3K is no longer effectively suppressed, and conditions favor self-renewal rather than differentiation. Second, at higher concentrations of LY 294002 (>50 μM), we saw a concentration-dependent increase in cell death consistent with a role for PI3K in cell survival. This suggests that different thresholds of PI3K signaling in hESCs are responsible for different cell fate outcomes (i.e., self-renewal and cell survival). Another parameter that is critical for normal DE differentiation is the exposure time for which hESCs are treated with LY 294002. If drug is removed at any time during a typical 4–5-day differentiation experiment, elevated levels of DE markers (SOX17, CXCR4, and GSC) are not observed (data not shown), indicating that sustained inhibition of PI3K is required. It is unclear whether this is due to a delay in differentiation or a blockade imposed by restoration of PI3K signaling.

Several lines of evidence indicate that we are specifying DE in our cultures and not selecting against non-DE cell types. First, the yield of DE (number of SOX17+ cells per number of plated hESCs) is typically >4.0 over 4 days, and second, only modest levels of cell death are seen in these cultures under the conditions used (data not shown). The amplification of cells in these cultures under DE differentiation conditions is consistent with our observations that >95% of LY 294002-treated cells at days 1–5 incorporate BrdU (data not shown). This raises the possibility that hESC-DE could be amplified once conditions are optimized.

TRK receptor signaling through PI3K signaling has recently been shown to promote survival of hESCs plated at clonal density under specific conditions [18]. Our findings describe a separate role for PI3K in hESC self-renewal in two ways. First, we show that PI3K promotes self-renewal by imposing a differentiation blockade that has an impact on formation of mesendoderm, definitive endoderm, and, potentially, mesoderm. Second, the role for PI3K that we define is dependent on signaling by insulin and/or IGF. PI3K therefore seems to be a critical pathway for maintenance of hESCs by having an impact on different aspects of hESC behavior. PI3K signaling has been implicated previously in murine ESC self-renewal as a consequence of LIF signaling [19, 22]. Future work will address the mechanism by which PI3K antagonizes lineage commitment.

The most obvious mechanism by which PI3K blocks differentiation would involve the inhibition of Smad2 signaling. It is unlikely that PI3K inhibits Smad signaling, however, for two reasons. First, Smad2 is activated in hESCs independently of PI3K signaling status [27, 28] (M. McLean, S. Dalton, unpublished data). Second, Activin target genes such as Nodal, Lefty A, and Lefty B are expressed in hESCs, and this is also independent of PI3K signaling status (our unpublished data). It is possible that there are PI3K-sensitive Activin response genes yet to be identified that could be critical for cell fate commitment. Although Smad2 is insensitive to PI3K signaling, perhaps Smad coactivators that bind to subsets of Activin target genes are the focus of this regulation.

In summary, we have developed a new method for the efficient generation of DE from hESCs. We define the two absolute requirements for DE formation from hESCs: Activin/Nodal signaling and low PI3K activity. Our observations explain why it is necessary to use low serum for Activin A to specify DE, but the mechanism by which PI3K antagonizes differentiation remains to be elucidated. Our findings underscore the role of PI3K signaling in hESC biology and, for the first time, in cell fate specification. Future work will address the mechanism by which PI3K blocks differentiation of hESCs in response to Activin A and Nodal.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

We thank members of the Dalton and Baetge laboratories for useful comments throughout the course of this work. We also thank Melissa Carpenter, Alberto Hayek, Lori Sussel, and Jim Wells for helpful comments and advice. This work was funded by grants from the Georgia Cancer Coalition and the Georgia Research Alliance (to S.D.).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
Dalton_Supplementary_Table_1.pdf7KSupplemental Table
Dalton_Supplemental_Figure_1.pdf90KSupplemental Figure 1
Dalton_Supplemental_Figure_2.pdf16KSupplemental Figure 2
Dalton_Supplemental_Figure_3.pdf17KSupplemental Figure 3
Dalton_Supplemental_Figure_4.pdf168KSupplemental Figure 4

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.